Abstract
A version of feature geometry in which the features are phonetic categories with a potential for distinctive function is presented and applied to the speech of a severely dysfluent adult, who frequently uses voiced nasal clicks to realize target sonorants, and occasionally non-nasal clicks to realize target obstruents. Many of the phonetic features of normal English are shown to be present in these realizations where they have the same kind of pattern of redundancy and distinctiveness found in normal realizations. The geometry is also seen to account for why certain phonemic oppositions, but not others, are lost when their members are realized as clicks. Involuntary dorsovelar occlusion seems to be a common characteristic of the speaker's dysfluency on target sonorants, and click realizations are explained as his response to this. By initiating a velaric ingressive air stream he can impose target buccal articulations while the system comprising the larynx, pharynx and nasal chamber is given over to the generation of sonority. This mode of realization is described as ‘disjunctive’. It is proposed that the buccal and extrabuccal geometries be autonomously associated with the same position in syntagmatic phonological structure.